There currently exists no functional model of speech perception in dysarthria that captures the critical interface between speech signal characteristics and the cognitive-perceptual processes brought to bear on that signal by the listener. Yet such a model is necessary, not only to explain intelligibility deficits, but also to guide and justify treatment decisions in clinical practice. Two series of experiments will be undertaken, which focus on the signal-listener interface for lexical segmentation, or the perceptual task of parsing the continuous acoustic stream into discrete words. The first set will focus on the nature of the intelligibility deficit by examining speech perception errors among different forms and severity levels of dysarthria. This will define and establish the relationships among segmental and suprasegmental deficit patterns, dysarthria severity levels, and the perceptual consequences of each. The second set of experiments will focus on the sources of intelligibility gains, by directly manipulating listener constraints in a training paradigm. In both series of experiments, predictions proposed by two accounts of lexical segmentation will be tested. These include the Metrical Segmentation Strategy Hypothesis (MSS; Cutler & Norris, 1988; Cutler & Butterfield, 1992), and the Hierarchical Model of Speech Segmentation (HMSS; Mattys, S.L. The hierarchical model of speech segmentation. BBSRC, 2003-2006). Lexical boundary error and segmental analyses will be conducted on listeners' transcription of phrases produced by speakers with different forms and severities of dysarthria. It is predicted that, for a given pattern of dysarthria (form), there will be evidence of differences in the effectiveness of listeners' cognitive-perceptual strategies, directly traceable to severity of speech deficit. For a given level of speech deficit severity, there will be evidence of differences in the effectiveness of listeners' cognitive-perceptual strategies, directly traceable to dysarthria form. By examining perceptual error patterns elicited in a training paradigm, it will be possible to identify which aspects of the acoustic signal are of perceptual salience in a default mode, and which features can be elevated in perceptual salience via training. Information learned about differences in the perceptual processing of different forms and severities of dysarthria will be used to develop a model of speech intelligibility deficits in dysarthria, and will have applicability to management programs in speech rehabilitation.